COP 4710 Database Systems Spring 2004 Day 27

COP 4710: Database Systems Spring 2004 -Day 27 – April 14, 2004 – Introduction to Data Mining Instructor : Mark Llewellyn markl@cs. ucf. edu CC 1 211, 823 -2790 http: //www. cs. ucf. edu/courses/cop 4710/spr 2004 School of Electrical Engineering and Computer Science University of Central Florida COP 4710: Database Systems (Day 27) Page 1 Mark Llewellyn ©

Three Dimensional View of Data Rubik’s Cube Horizontal slice Location Horizontal slice. All products at one location over all dates. te a D Product COP 4710: Database Systems (Day 27) Page 2 Mark Llewellyn ©

Three Dimensional View of Data (cont. ) Vertical slice Location Vertical slice. One product at all locations over all dates. te a D Product COP 4710: Database Systems (Day 27) Page 3 Mark Llewellyn ©

Three Dimensional View of Data Vertical slice Location Horizontal slice te a D Product COP 4710: Database Systems (Day 27) Horizontal slice, vertical dice. One product at one location over all dates. Vertical dice of a horizontal slice Page 4 Mark Llewellyn ©

Three Dimensional View of Data Vertical slice Horizontal slice Location Intersection of a horizontal slice and vertical slice yields all products at one location on one date. te a D Intersection of a horizontal slice and a vertical slice Product COP 4710: Database Systems (Day 27) Page 5 Mark Llewellyn ©

Three Dimensional View of Data Vertical slice Horizontal slice Location Intersection of a horizontal slice and vertical slice yields all products at one location on one date. te a D Intersection of a horizontal slice and a vertical slice Product COP 4710: Database Systems (Day 27) Page 6 Mark Llewellyn ©

Three Dimensional View of Data Vertical slice Horizontal slice Location Sliced and diced. One product at one location on one date. te a D Sliced and diced. Product COP 4710: Database Systems (Day 27) Page 7 Mark Llewellyn ©

Three Dimensional View of Data Location Sliced and diced. One product at one location on one date. te a D Product COP 4710: Database Systems (Day 27) Page 8 Mark Llewellyn ©

Three Dimensional View of Data Location Sliced and diced. One product at one location on one date. te a D Product COP 4710: Database Systems (Day 27) Page 9 Mark Llewellyn ©

Introduction to Data Mining • The amount of data maintained in computer files and databases is growing at a phenomenal rate. • At the same time, the users of these data are expecting more sophisticated information from them. – A marketing manager is no longer satisfied with a simple listing of marketing contacts, but wants detailed information about customers’ past purchases as well as predictions of future purchases. • Simple structured/query language queries are not adequate to support these increased demands for information. • Data mining has evolved as a technique to support these increased demands for information. COP 4710: Database Systems (Day 27) Page 10 Mark Llewellyn ©

Introduction to Data Mining (cont. ) • Data mining is often defined as finding hidden information in a database. • Alternatively, it has been called exploratory data analysis, data driven discovery, and deductive learning. • We’ll look at a somewhat more focused definition that was provided by Simoudis (1996, IEEE Expert, Oct, 26 -33) who defines data mining as: The process of extracting valid, previously unknown, comprehensible, and actionable information from large database and using that information to make crucial business decisions. COP 4710: Database Systems (Day 27) Page 11 Mark Llewellyn ©

Introduction to Data Mining (cont. ) • Traditional database queries access a database using a well-defined query state in a language such as SQL. The output of the query consists of the data from the database that satisfies the query. The output is usually a subset of the database, but it may also be an extracted view or contain aggregations. • Data mining access of the database differs from this traditional access in three major areas: 1. Query: The query might not be well formed or precisely stated. The data miner might not even be exactly sure of what they want to see. 2. Data: The data access is usually a different version from that of the operational database (it typically comes from a data warehouse). The data must be cleansed and modified to better support mining operations. 3. Output: The output of the data mining query probably is not a subset of the database. Instead it is the output of some analysis of the contents of the database. COP 4710: Database Systems (Day 27) Page 12 Mark Llewellyn ©

Introduction to Data Mining (cont. ) • The current state of the art in data mining is similar to that of database query processing in the late 1960 s and early 1970 s. Over the next decade or so, there will undoubtedly be great strides in extending the state of the art with respect to data mining. • We will probably see the development of “query processing” models, standards, and algorithms targeting data mining applications. • In all likelihood we will also see new data structures designed for the storage of database being using specifically for data mining operations. • Although data mining is still a relatively young discipline, the last decade has witnessed a proliferation of mining algorithms, applications, and algorithmic approaches to mining. COP 4710: Database Systems (Day 27) Page 13 Mark Llewellyn ©

A Brief Data Mining Example • Credit card companies must determine whether to authorize credit card purchases. Suppose that based on past historical information about purchases, each purchase is placed into one of four classes: (1) authorized, (2) ask for further identification before authorization, (3) do not authorize, and (4) do not authorize and contact the police. • The data mining functions here are twofold. • – First, the historical data must be examined to determine how the data fit into the four classes. That is, how all of the previous credit card purchases should be classified. – Second, once classified the problem is to apply this model to each new purchase. The second step above can be stated as a simple database query if things are properly set-up, the first problem cannot be solved with a simple query. COP 4710: Database Systems (Day 27) Page 14 Mark Llewellyn ©

Introduction to Data Mining (cont. ) • Data mining involves many different algorithms to accomplish different tasks. All of these algorithms attempt to fit a model to the data. • The algorithms examine the data and determine a model that is the closest fit to the characteristics of the data being examined. • Data mining algorithms can be viewed as consisting of three main parts: 1. Model: The purpose of the algorithms is to fit a model to the data. 2. Preference: Some criteria must be used to fit one model over another. 3. Search: All algorithms require some technique to search the data. COP 4710: Database Systems (Day 27) Page 15 Mark Llewellyn ©

Data Mining Models • A predictive model makes a prediction about values of data using known results found from different data. Predictive modeling is commonly based on the use of other historical data. – For example, a credit card use might be refused not because of the user’s own credit history, but because a current purchase is similar to earlier purchases that were subsequently found to be made with stolen cards. – Predictive model data mining tasks include classification, regression, time series analysis, and prediction (as a specific data mining function). COP 4710: Database Systems (Day 27) Page 16 Mark Llewellyn ©

Data Mining Models (cont. ) • A descriptive model identifies patterns or relationships in data. Unlike the predictive model, a descriptive model serves as a way to explore the properties of the data examined, not to predict new properties. – For example, a credit card purchase may be not authorized because the amount of the charge is way out of line with your typical charges. In other words, if you have a past history where your average charge amount is $100. 00 and the current transaction is for $5000. 00 the charge might not be authorized using this model. This is a summarization technique. – Clustering, summarizations, association rules, and sequence discovery are usually viewed as descriptive in nature. COP 4710: Database Systems (Day 27) Page 17 Mark Llewellyn ©

Data Mining Models and Tasks Data Mining Descriptive Models Predictive Models Classification Prediction Regression Time-series Analysis Clustering Sequence Discovery Summarization Association Rules Data mining models and some typical tasks. Not an exhaustive listing. Combinations of these tasks yield more sophisticated mining operations. COP 4710: Database Systems (Day 27) Page 18 Mark Llewellyn ©

Basic Data Mining Tasks Classification (predictive model) • Classification maps data into predefined groups or classes. It is often referred to as supervised learning because the classes are not determined before examining the data. • Two examples of classification applications are determining whether to make a bank loan and identifying credit risks. • Classification algorithms require that the classes be defined based on data attribute values. They often describe these classes by looking at the characteristics of data already known to belong to the classes. – Supervised learning normally consists of two phases: training and testing. Training builds a model using a large sample of historical data called a training set, while testing involves trying out the model on new, previously unseen data to determine its accuracy and physical performance characteristics. COP 4710: Database Systems (Day 27) Page 19 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Classification (cont. ) • Pattern recognition is a type of classification where an input pattern is classified into one of several classes based on its similarity to these predefined classes. • The example on page 6 is an example of a general classification problem. • An example of pattern recognition would be an airport security system used to determine if passengers are potential terrorists or criminals. Each passenger’s face is scanned and its basic pattern (distance between eyes, size and shape of mouth, shape of head, etc. ) is identified. This pattern is compared to entries in a database to see if it matches any patterns that are associated with known offenders. COP 4710: Database Systems (Day 27) Page 20 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Classification (cont. ) • There are two major types of classification algorithms: tree induction and neural induction. • To illustrate the differences and similarity in these two techniques, consider the following example: – Suppose that we are interested in predicting whether a customer who is currently renting property is likely to be interested in buying property. – Assume that a predictive model has determined that only two variables are of interest: the length of time the customer has rented property and the age of the customer. – Tree induction presents the analysis in an intuitive way, using a decision tree (similar in some ways to a flow chart). A possible classification using tree induction is shown in the following diagram: COP 4710: Database Systems (Day 27) Page 21 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Classification (cont. ) Customer renting property more than 2 years? Yes No Customer over 25 years old Rent property Yes No This predictive model will classify customers into one of two categories: renters and buyers. The model will predict that customers who are over 25 years old and have rented for more than 2 years will buy property, others will rent. Rent property Buy property Classification Using An Induction Tree COP 4710: Database Systems (Day 27) Page 22 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Classification (cont. ) • Using neural induction, for the same example, would require the use of a neural network. A neural network contains collections of connected nodes with input, output, and processing at each node. Between the visible input and output layers may be a number of hidden processing layers. Each processing unit (the circles in the diagram) in one layer is connected to each processing unit in the next layer by a weighted value, expressing the strength of the relationship. The network attempts to mirror the way the human brain works in recognizing patterns by arithmetically combining all the variables associated with a given data point. In this way, it is possible to develop nonlinear predictive models that “learn” by studying combinations of variables and how different combinations of variables affect different data sets. neural network 0. 5 customers who have rented 2 or more years 0. 7 0. 4 classification: rent or buy 0. 6 customers who are 25 years old or more COP 4710: Database Systems (Day 27) 0. 4 hidden processing layer 0. 3 Page 23 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Regression (predictive model) • Regression is used to map a data item to a real valued prediction variable. • In actuality, regression involves the learning of the function that does this mapping. • Regression assumes that the target data fit into some known type of function (i. e. , linear, logistic, etc. ) and then determines the best function of this type that models the given data. • Some type of error analysis is used to determine which function is “best”, i. e. , produces the least total error. • As an example of simple linear regression let’s suppose that you are maintaining a retirement savings portfolio and wish to reach a certain level of savings before retirement. Periodically, you will predict what your savings will be based on the current amount and several past amounts. Using a simple linear regression formula you then predict what the value will be in the future by fitting the past values to a linear function and then use that function to predict values at points in the future. Based on these values, you then alter (or not) your investment portfolio. COP 4710: Database Systems (Day 27) Page 24 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Regression (cont. ) • Linear regression attempts to fit a straight line through the plot of the data, such that the line is the best representation of the average of all observations at that point in the plot. • The problem with linear regression is that the technique only works well with linear data and is sensitive to the presence of outliers (data values which do not conform to the expected norm). • Although nonlinear regression avoids the main problems of linear regression, it is still not flexible enough to handle all possible shapes of the data plot. • This is where the traditional statistical analysis methods and data mining methods begin to diverge. Statistical measurements are fine for building linear models that describe predictable data points, however, most data is not linear in nature. • Data mining requires statistical methods that can accommodate nonlinearity, outliers, and non-numeric data. COP 4710: Database Systems (Day 27) Page 25 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Time Series Analysis (predictive model) • With time series analysis, the value of an attribute is examined as it varies over time. The values usually are obtained as evenly spaced time points (daily, weekly, hourly, etc. ). • A time series plot is used to visualize the time series. In the example below, the plots for Y and Z appear to have similar behavior, while X appears less similar. X Y Z COP 4710: Database Systems (Day 27) Page 26 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Time Series Analysis (cont. ) • There are three basic functions performed in time series analysis. • In one case, distance measures are used to determine the similarity between different time series. For example, using the time series on the previous page we could look at the difference in daily stock prices between the three companies, or perhaps the difference between their beginning and ending prices, etc. . • In the second case, the structure of the line is examined to determine (and perhaps classify) its behavior. This could be a generality, such as X appears to be trending upwards, or it could use very specific curve fitting techniques. • A third case would occur when historical time series plots are used to predict future values. Various extrapolation techniques could be applied. COP 4710: Database Systems (Day 27) Page 27 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Time Series Analysis (cont. ) • As an example of how to use time series analysis, suppose that you are deciding whether to purchase stock from Companies X, Y, or Z. Assuming that the time series plots illustrated on page 14 were tracking the daily stock prices for each company, you might decide to purchase stock in either Y or Z because they appear to be less volatile (fluctuate less on a daily basis) that does the stock for company X. On the other hand you might decide to purchase stock in company X because it shows an overall growth which is larger than either of the other two stocks. COP 4710: Database Systems (Day 27) Page 28 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Prediction (predictive model) • Many real-world data mining application can be seen as predicting future data states based on past and current data. • Prediction can be also be viewed as a type of classification. Note that this is a data mining task which is different from the prediction model, although the prediction task is a type of the prediction model. The difference is that prediction is predicting a future state rather than a current state. • An example of prediction can be illustrated with the application of the prediction of flooding. In general predicting flooding is a difficult problem. One approach uses monitors placed at various points along a river. The monitors collect data relevant to flood prediction such as water levels, rain amounts, time, humidity, etc. . Then the water level at a potential flooding point in the river can be predicted based on the data collected by the sensors upriver from this point. The prediction must be made with respect to the time the data were collected. COP 4710: Database Systems (Day 27) Page 29 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Clustering (descriptive model) • Clustering is similar to classification except that the groups are not predefined, but rather defined by the data alone. • Clustering is alternatively referred to as unsupervised learning or segmentation (actually, segmentation is a special case of clustering although many people refer to them synonymously). • Clustering can be thought of as partitioning or segmenting the data into groups that might or might not be disjoint. • Clustering is usually accomplished by determining the similarity among the data on predefined attributes. The most similar data are grouped into clusters. • Since clusters are not predefined, a domain expert is often required to interpret the meaning of the created clusters. • As an example of clustering, suppose that you are an instructor for COP 3502 and you have 10 different lab sections for the course. Students attend a particular lab section. If you have a database in which each student’s lab quiz scores are recorded, then you can cluster (segment) the database using the lab section as a clustering attribute and cluster students attending the same lab section together. COP 4710: Database Systems (Day 27) Page 30 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Summarization (descriptive model) • Summarization maps data into subsets with associated simple descriptions. It extracts or derives representative information about the database. • This is commonly accomplished by actually retrieving portions of the data. Alternatively, summary type information (e. g. , the mean of some numeric attribute) can be derived from the data. • Summarization succinctly characterizes the contents of the database. • Summarization is also called characterization or generalization. • An example of summarization would be one of the many criteria used to compare universities by U. S. News and World Report which is average SAT score. This summarization is used to estimate the type and intellectual level of a student body. COP 4710: Database Systems (Day 27) Page 31 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Association Rules (descriptive model) • Association is also called link analysis or affinity analysis, and refers to the data mining task of uncovering relationships among the data. • The best example of this type of application is to determine association rules. An association rule is a model that identifies specific types of data associations. These associations are often used in the retail sales world to identify items that are frequently purchased together. This is commonly referred to as market basket analysis. – • As an example of association rules, suppose that a grocery store manager is trying to decide whether or put bread on sale. To help determine the impact of this decision, the manager generates association rules that show what other products are frequently purchased with bread. Suppose that they discover that 60% of the time bread is purchased with pretzels and 70% of the time bread is purchased with jelly. Based on these facts, the manager attempts to capitalize on the association between bread, pretzels and jelly by placing some pretzels and jelly on the end of the aisle where the bread is located. In addition, he decides never to place both of these items on sale at the same time! Associations are also used in many other applications such as predicting the failure of telecommunication switches. COP 4710: Database Systems (Day 27) Page 32 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Association Rules (cont. ) • When using association rules, one must remember that these are not casual relationships. They doe not represent and relationship inherent in the actual data as is the case with functional dependencies, or in the real world. • There is probably no relationship between bread pretzels that causes them to be purchased together. Furthermore, there is no guarantee that this association will apply in the future. • However, association rules are heavily used in the retail sector in creating effective advertising, marketing and inventory control strategies. COP 4710: Database Systems (Day 27) Page 33 Mark Llewellyn ©

Basic Data Mining Tasks (cont. ) Sequence Discovery (descriptive model) • Sequential analysis or sequence discovery is used to determine sequential patterns in data. These patterns are based on a time sequence of actions. • These patterns are similar to associations in that the data (or events) are found to be related, but the relationship is based on time. This is different from market basket analysis, which requires the related objects to be purchased at the same time. In sequence discovery, the items are purchased over some period of time in some order. • For example, most people who purchase a DVD player may be found to purchase DVDs within one week. • Temporal association rules really fall into this category although some people try to force the issue and maintain them as strict association rules. COP 4710: Database Systems (Day 27) Page 34 Mark Llewellyn ©

Knowledge Discovery in Databases vs. Data Mining • The terms knowledge discovery in databases (KDD) and data mining are often used interchangeably. However, over the last few years KDD has been used to refer to a process consisting of many steps, while data mining is only one of these steps. • Data mining has become a specific operation in the wider arena of knowledge discovery. Knowledge discovery in databases (KDD) is the process of finding useful information and patterns in data. Data mining is the use of algorithms to extract the information and patterns derived by the KDD process. • KDD is a process that involved many different steps. The input to this process is the data and the output is the useful information desired by the users. However, the objective may be unclear or inexact. The process itself is interactive and may require much elapsed time. • To ensure the accuracy and usefulness of the results, interaction throughout the process with both domain experts and technical experts may be needed. COP 4710: Database Systems (Day 27) Page 35 Mark Llewellyn ©

The KDD Process • The KDD process consists of the following five basic steps: 1. Selection: The data needed for the data mining process is obtained from many different and heterogeneous data sources. 2. Preprocessing: The data to be used by the process may have incorrect or missing data. There may be anomalous data from multiple sources involving different data types and metrics. There may be many different activities performed during this step. Erroneous data may be corrected or removed, whereas missing data must be supplied or predicted (often using data mining tools). 3. Transformation: Data from different sources must be converted into a common format for processing. Some data may be encoded or transformed into more usable formats. Data reduction may be used to reduce the number of possible data values being considered. 4. Data mining: Based on the data mining task being performed, this step applies the algorithms to the transformed data to generate the desired results. 5. Interpretation/evaluation: How the data mining results are presented to the users is extremely important because the usefulness of the results is dependent on it. Various visualization and GUI strategies are used in this last step. COP 4710: Database Systems (Day 27) Page 36 Mark Llewellyn ©

Data Mining Issues • There are many important implementation issues associated with data mining: 1. Human interaction: Since data mining problems are often not precisely stated, interfaces may be needed with both domain and technical experts. Technical experts are used to formulate the queries and assist in interpreting the results. Users must identify training data and desired results. 2. Overfitting: When a model is generated that is associated with a given database state, it is desirable that the model also fit future database states. Overfitting occurs when the model does not fit future states. This may be caused by assumptions that are made about the data or may simply be caused by the small size of the training database. For example, a classification model for an employee database may be developed to classify employees as short, medium, or tall. If the training database is quite small, the model might erroneously indicate that a short person is anyone under 5’ 8” because there is only one entry in the training database under 5’ 8”. In this case, many future employees would be erroneously classified as short. Overfitting can arise under other circumstances as well, even though the data are not changing. COP 4710: Database Systems (Day 27) Page 37 Mark Llewellyn ©

Data Mining Issues (cont. ) 3. Outliers: There are often many data entries that do not fit nicely into the derived model. This becomes even more of an issue with VLDBs. If a model is developed that includes these outliers, then the model may not behave well for data that are not outliers. 4. Interpretation of results: Currently, data mining output may require experts to correctly interpret the results, which might otherwise be meaningless to the average database user. 5. Visualization of the results: To easily view and understand the output of data mining algorithms, visualization of the results is essential. Selection of the appropriate tool becomes critical to aid in the interpretation. 6. Large datasets: The massive datasets associated with data mining create problems when applying algorithms designed for small datasets. Many modeling applications grow exponentially on the dataset size and thus are too inefficient for larger datasets. Sampling and parallelization are effective tools to attack this scalability problem. COP 4710: Database Systems (Day 27) Page 38 Mark Llewellyn ©

Data Mining Issues (cont. ) 7. High dimensionality: A conventional database schema may be composed of many different attributes. The problem here is that not all attributes may be needed to solve a given data mining problem. In fact, the use of some attributes may interfere with the correct completion of a data mining task. The use of other attributes may simply increase the overall complexity and decrease the efficiency of an algorithm. This problem is sometimes referred to as the dimensionality curse, meaning that there are many attributes (dimensions) involved and it is difficult to determine which ones should be used. One solution to this high dimensionality problem is to reduce the number of attributes, which in known as dimensionality reduction. However, determining which attributes are not needed is not always easy to do. 8. Multimedia data: Most previous data mining algorithms are targeted to traditional data types (numeric, character, text, etc. ). The use of multimedia data such as found in GIS databases complicates or invalidates many proposed algorithms. COP 4710: Database Systems (Day 27) Page 39 Mark Llewellyn ©

Data Mining Issues (cont. ) 9. Missing data: During the preprocessing phase of KDD, missing data may be replaced with estimates. This and other approaches to handling missing data can lead to invalid results in the data mining step. 10. Irrelevant data: Some attributes in the database might not be of interest to the data mining task being developed. 11. Noisy data: Some attribute values might be invalid or incorrect. These values are often corrected before running data mining applications. 12. Changing data: Databases cannot be assumed to be static. However, most data mining algorithms do assume a static database. This requires that the algorithms be completely rerun anytime the database changes. 13. Integration: The KDD process is not currently integrated into normal data processing activities. KDD requests may be treated as special, unusual, or one-time needs. This makes them inefficient, ineffective and not general enough to be used on an ongoing basis. Integration of data mining functions into traditional DBMSs is certainly a desirable goal. 14. Application: Determining the intended use for the information obtained from the data mining function is a challenge. How business executives can effectively use the output is sometimes considered the more difficult part, not the running of the algorithms themselves. Because the data are of a type that has not previously been known, business practices may have to be modified to determine how to effectively use the information uncovered. COP 4710: Database Systems (Day 27) Page 40 Mark Llewellyn ©